none reviewed yet · 4 benchmark observations · latest capture Sep 8, 2026
Inspect the evidenceZ.ai · MODEL PROFILE
GLM-5.2
Teams that want an MIT-licensed, self-hostable model with good design taste and instant responses
Saved editorial fit for your current task lens; not a benchmark result
How scoring worksVerified rates (Sep 8, 2026) · 100 runs / month at your token settings
Adjust workloadAt a glance
9 facts- Provider
- Z.ai
- Type
- API model
- Access
- Z.ai API, GLM Coding Plan, Alibaba Cloud Model Studio (multi-region), Hugging Face weights
- Context
- 1,000,000 tokens
- Maximum output
- n.a. on fetched pages
- License / weights
- Open weights
- Input modalities
- text
- Output modalities
- text
- Pricing notes
- $1.40 in / $4.40 out per 1M with $0.26 cached input (Z.ai); Alibaba resale $1.40/$4.40 in Singapore and $1.10/$3.851 in China, Germany, US and Hong Kong; `glm-5.2-us` $1.40/$4.4
The practical trade-offs
Documented strengths
- Design Arena Website rank 8 (1315) - best open-weight design-preference score on that board
- MIT licence with open weights and a 1M context
- TTFT 1.60 s at 71.1 tok/s makes it feel immediate
- SWE-bench Pro 62.1% at $1.40/$4.40 (or $1.10/$3.851 in several regions)
Limitations to check
- GLM-5.3 costs the same and is much stronger (DeepSWE 46.2 -> 66.9)
- DeepSWE 46.2 and Terminal Bench 3.0 4.6 show weak long-horizon agentic coding
- Text-only on the evidence available
- AA non-reasoning row collapses to 26
Follow the source
8 source linksEvidence & conditions
8Source links
NoneReviewed records
4Benchmark runs
Sep 8, 2026Latest capture
- Evidence source 1
docs.z.aiDate not recorded - Evidence source 2
alibabacloud.comDate not recorded - Evidence source 3
docs.qwencloud.comDate not recorded - Evidence source 4
arena.aiDate not recorded - Evidence source 5
benchlm.aiDate not recorded - Evidence source 6
benchlm.aiDate not recorded - Evidence source 7
artificialanalysis.aiDate not recorded - Evidence source 8
huggingface.coDate not recorded
Benchmark observations
Terminal-Bench 2.1 (Terminus 2) Inspect conditions
- name
- Terminal-Bench 2.1 (Terminus 2)
- value
- 81
- unit
- %
- source Url
- https://huggingface.co/zai-org/GLM-5.2
- checked At
- 2026-09-08T19:28:18Z
- attribution
- provider-reported
- configuration
- GLM-5.2; Terminus 2; JSON parser; four-hour limit; temperature=1; top_p=1; 48K maximum output; 500 episodes; 256K context; 4 CPU / 8 GB.
Terminal-Bench 2.1 (Claude Code) Inspect conditions
- name
- Terminal-Bench 2.1 (Claude Code)
- value
- 82.7
- unit
- %
- source Url
- https://huggingface.co/zai-org/GLM-5.2
- checked At
- 2026-09-08T19:28:18Z
- attribution
- provider-reported
- configuration
- GLM-5.2; Claude Code 2.1.167; temperature=1; top_p=0.95; 131072 maximum output via proxy override; no wall-clock limit; average of five runs.
DeepSWE Inspect conditions
- name
- DeepSWE
- value
- 46.2
- unit
- %
- source Url
- https://huggingface.co/zai-org/GLM-5.2
- checked At
- 2026-09-08T19:28:18Z
- attribution
- provider-reported
- configuration
- GLM-5.2; official pier mini-SWE framework; temperature=1; top_p=1; two-hour limit; 400K context; 2 CPU / 8 GB; no internet.
Arena WebDev Frontend Inspect conditions
- name
- Arena WebDev Frontend
- value
- 1594
- unit
- arena_score
- source Url
- https://arena.ai/leaderboard/code/webdev/frontend
- checked At
- 2026-09-08T19:28:18Z
- attribution
- benchmark-publisher
- configuration
- Exact row glm-5.2-max; source results dated 2026-09-05; human preference; confidence interval +/-8; 8056 votes. This is a checked source snapshot, not a newly run September 8 evaluation.
Source reports, measured results, and editorial assessments are separate evidence types. Source dates do not imply a new benchmark run. Capture dates are shown when recorded; otherwise we say the capture date was not recorded. Recorded dates are compared with this catalogue snapshot (Sep 11, 2026), not today’s date.
Method AIS-V2
How this page is assessed
Quality claims follow Method AIS-V2.
Money never buys rank — commercial relationships cannot change rank, eligibility, evidence quality, or publication timing.
Submit or claim a listing is free and does not skip the editorial queue.
Private · this device only